Design of a Spherical Robot Arm with the Spiral Zipper Prismatic Joint

Foster Collins and Mark Yim

Abstract— A novel prismatic joint called a Spiral Zipperis used to create a 3DOF robot arm in a spherical robotconfiguration. The Spiral Zipper can be very compact as ithas a large extension to compression ratio. An initial prototypehas shown a ratio of over 14:1. The Spiral Zipper is very strongin compression, but maybe loose under tension and moments.A tether based system ensures the prismatic joint is alwaysin compression while enabling spherical coordinate positioningwith a long reach, high force, low mass design. While havingtypically an order magnitude higher strength and lower weightratio for a given reach compared to standard industrial robotarms, the arm is slower and was not designed for high precision.These characteristics may be applicable for mounting on mobilerobots or flying vehicles. This paper introduces the design andtesting of several prototypes.

I. INTRODUCTION

Robot arms have been used in industry for many decades.More recently researchers have been exploring the use ofrobot arms mounted on mobile bases [1] and even flyingvehicles [2]. One major difference over those mounted onfixed bases is that the mass of the arm is much moreimportant especially if the arm has a long reach which canlead to tip-over conditions for mobile platforms.

Robot arm systems can be classified by many differentmetrics, but two important ones are the working load andvolume of reachable workspace. While an arm capable of fulldextrous workspace requires six degrees of freedom (DOF),reachable workspace considers three translation DOF withoutorientation. Common configurations for these three DOFinclude a serial arrangement of prismatic (P) and revolute(R) joints such as: articulated (RRR), cartesian or gantry(PPP), cylindrical (PPR), spherical also called polar (RRP)and SCARA (RRP) where the two revolute joints are parallelto each other [3].

One advantage of spherical configurations over others isthat collision avoidance and reachability concerns are easier.The standard human-like articulated arm often has difficultyin cluttered environments because the links and elbow sweepout a volume colliding with the environment even whenthe path of the end-effector or carried object does not. Thestandard solution is to add more DOF so the elbow can bemaneuvered into places that are obstacle-free along with thecarried object. A simple approach for path planning withhyper-redundant snake-like arms is to approximate the body-follows-head paths in which the swept volume is only that ofthe carried object [4]. With line of sight, the shortest body-follows-head path would be a straight line and requires only

The authors are with the Dept. of Mech. Eng. & Appl. Mechanicsand GRASP Lab at the University of Pennsylvania, Philadelphia, PA.fcollins@seas.upenn.edu, yim@grasp.upenn.edu

one DOF. A guaranteed collision-free path from one positionto another then has three motions: a retraction, a reorienting,and an extension - ideal for a spherical configuration.

For robots doing human-like tasks it is useful to see howhumans perform. The vast majority of objects manipulatedby humans are reachable by a straight line, even in a clutteredenvironment. If you can see the object, there is line-of-sightand therefore a straight line path to that object. It is rarefor humans to grasp objects that they cannot see. Examplesof specialized exceptions include reaching into pockets orreaching blindly around furniture or under a bed.

The goal characteristics of this system are long reach,high-strength to weight ratio, low profile and low cost.Precision and speed are not a priority. The core of thissystem is a novel prismatic joint called the Spiral Zipper[5]. Important metrics that characterize a prismatic jointinclude, the extension ratio (length fully extended to lengthfully collapsed), the strength of the formed tube (abilityto withstand forces) and the mass. Often the latter twoare coupled so a strength to weight ratio is often used tocharacterize both.

This paper will discuss the design and testing of a newrobot arm system using the Spiral Zipper in conjunction withwinches and cables shown in Figure 1 able to achieve a highextension ratio as well as forming a high strength to weightratio column to support large loads in a spherical manipulatorsystem.

Fig. 1. A 17mm diameter Spiral Zipper mechanism is mounted on a gimbalwith 2 winch/cables in a spherical robot arm configuration.

A. Related WorkThere are three forms of related work: spherical robots,

cable-based systems and expanding tube prismatic joints.Spherical robot arm systems include the Stanford arm,

and the first Unimate from the 1960’s [3]. The prismatic joint

This accepted article to ICRA 2016 is made available by the authors in compliance with IEEE policy.Please find the final, published version in IEEE Xplore, DOI: 10.1109/URAI.2016.7487363.

c©2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any currentor future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collectiveworks, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

used in these systems are rack-and-pinion style or leadscrewmechanisms where a long rigid link is translated to extendor retract the end effector. Even when not extended forward,the unused portion of the link is extended backward requiringlarge clearance. These system precision requirements oftenlead to stiff links which are higher cost and/or heavy. Thelarge clearance, high cost and weight of these systemsmay be the reason that spherical robot arms are not aspopular as articulated systems. With the growing popularityof lightweight human-safe industrial arms such as Baxterand Universal Robots UR3 there has been a trend towardslower precision, lower inertia, lower cost systems capable ofworking in the presence of humans.

Winch and cable actuation is ideal when the goals for arobot arm are long reach, light weight and low profile, espe-cially when speed and precision are not important. Comparedto other transmissions with rigid links, the range of motioncan be arbitrarily long (e.g. 100’s of meters) winches can bevery strong if slow, and the mass is negligible when consid-ering the length of actuation. The major caveat is that thewinch cables are unidirectional. Cable-based robot systemshave also been in existence for several decades, most notablythe NIST robocrane [6] and the Skycam video systems nowpopular in sports arenas. However, these systems use justwinches and cables and rely on a large frame to define theworkspace of the robot, where as in our case, a prismaticjoint is used in conjunction with the cable system, resultingin a much lighter weight system for a given workspace.

The Spiral Zipper includes a band of material that canjoin to itself helically to form a tube. The oldest similarconcept comes from a 1939 patent for a lifting jack [7].More recently, the Zippermast [8] and the Spiralift [9] aretwo commercially available systems that work similarly.

The Zippermast developed by Geo Systems Incorporatedis a mechanism that uses three bands of metal that join in atriangle. The bands unspool from three separate coils in thebase and interlock teeth to form an ad hoc beam. They comein different sizes, but the largest extends 7.5m and supports25kg [9]. It is primarily used for an extension mast to get ahigh vantage point for sensors or antennae on a small groundrobot [10]. This is a good example of a large extension ratio,but a beam with a triangle cross section is not as strong asa cylinder all else being equal.

The Spiralift developed by the company Gala Systems,is a custom theater stage lift. It employs a system of twosteel bands, which form a rigid circular column. The first isa helical coil with teeth cut into the outside edge. The secondband is thin steel with perforated edges, wound similarly asa photographic filmstrip [8]. By wrapping the band aroundthe coil with the inner teeth locking the band together, itforms a very rigid column. The downside of this design iscomplexity and cost.

In contrast to these recent developments the proposeddesign requires only a single piece band and combines thebasic idea of an expanding tube into a 3DOF positioningsystem with high strength to weight ratio. Whereas theprevious are both stiffer and stronger, likely capable of higher

precision, precision is not considered a priority in this case.That said, using these systems in place of the Spiral Zipperwould still yield useful 3DOF positioning systems.

II. OPERATING PRINCIPLE

The basis of the Spiral Zipper design is a long, thin bandwith mating teeth along both edges. A rotating mechanismwe call a ”slider” wraps the band into a helix while meshingthe teeth on the bottom edge of one wrap with the teeth on thetop of the wrap below. This helixed column is theoreticallyonly limited by the length of the band.

The resulting column has exceptional performance incompression, but because of the zipping action, tension andmoments applied to the column result in significant playespecially on a long column. This Spiral Zipper column canbe combined with a cable system to exploit the exceptionalperformance in compression. Cables on winches betweena base and the distal end of the column as shown inFigure 1 can be used to position the end point while ensuringthe column remains in compassion under nominal loadingconditions.

The system can be broken into two major subsystems, theSpiral Zipper prismatic joint and the tether-gimbal rotationalDOF sub-system.

A. Spiral Zipper

The main components of the Spiral Zipper are the band,the slider and driving mechanism.

1) The Band: A key element to the viability of the tubeis formation of this helix. The geometry of the band in thishelix defines many of the important performance metrics ofthe mechanism. We define the following:

HB = Band HeightHT = Tooth HeightβI = Helix Incline AnglePB = Band Helix PitchD = Band DiameterSB = Arc Length of One WrapN = Number of teeth in a single Wrap (positive integervalue)PT = Band Tooth PitchH = Overall Column HeightLB = Length of Band

Fig. 2. A band laid out flat showing the dimensions of one wrap. Onlyone tooth is shown on the top and bottom edge.

Figure 2 shows a single unwrapped turn of the band.Assuming that the diameter and band and tooth heights have

been chosen, the helix angle βI can be calculated as follows:

βI = sin−1 HB −HT

πD(1)

And it follows that the pitch can be calculated by:

PB =HB −HT

cosβI(2)

The tooth pitch can be set.

PT = SB/(N) (3)

Extension height can be determined:

SB =√

(PB)2 + (D)2 (4)

H = LB/SBPB (5)

The above equation assumes that the band has the maletooth positioned directly above its female mating slot whenthe band is at the pitch angle, as shown in Figure 2. Depend-ing on the number of teeth desired, there is a quantized setof diameters that will work for a band.

As the band is wrapped, in order to maintain the structure,the top and bottom edges must not move as if they are rigidlyconnected. However, this connection must not be permanentallowing the wrapping and unwrapping process. Zippers dothis by (dis)engaging teeth that are (dis)assembled througha motion local to an edge element typically perpendicularto the direction of nominal load. Each element in a closedzipper prevents this local motion of the element further downthe chain, thus keeping the whole chain closed up to the endof the chain.

In the Spiral Zipper case, each tooth acts in the sameway. The teeth are engaged by wrapping, which is a purelyradial motion local to the tooth. Once engaged, for teethto become disengaged, a radial motion is required. For anytooth not near the local wrap, a radial motion is coupledwith a circumferential change as governed by Equation 4.This circumferential change cannot occur as the teeth preventsliding circumferentially. This design results in a the bandgeometry that is mostly planar easing manufacture (explainedin more detail in Section III-A).

2) Slider: The main purpose of the slider is the(dis)engagement of the teeth in the top and bottom edgeswhen forming the tube as described above, just as a the slideron a clothing zipper is the part where the two seams arejoined simply by pushing the two parts through the slider.

An exploded CAD drawing of a slider is shown in Fig-ure 3. The band enters the slider in the structure on the leftside of the figure and the tube grows out of the top. The bandonly needs to make contact with the slider in two parts, a lipat the bottom of the core that forms a helical ledge on whichthe band rests and the portion that joins the top and bottomedge (e.g. being squeezed by the slider body and the slideraccess door). Also just like a zipper, no other support otherthan the top of the tube and the slider portion is required tomaintain the structure.

Fig. 3. A slider mechanism exploded view.

3) Driving mechanism: Forming the tube can be viewedin one of two ways depending on view point: by feeding theband into a non-moving slider so the tube spins as it grows,or a spinning slider, winding the band around a growing tubethat does not spin. The spinning tube in the former case canmake it awkward for end effectors or tethers mounted to thedistal end. The latter case pushes complexity into the sliderwhich is the approach in this design, however it is oftenconceptually easier to consider the non-moving slider.

Unlike a clothing zipper, the band must be pushed throughthe slider automatically. We do this with a band drive gearpushing on slots laser cut into the band. Conceptually, amotor mounted to the slider can drive this gear to drive theband into or out of the slider, extending or collapsing thetube. To prevent the tube spinning, we instead drive the slideraround the growing tube while simultaneously driving theband in the opposite direction. This is achieved with gearingas illustrated in Figure 4. The band drive gear directly belowthe slider drives the band and simultaneously meshes with agear of the same pitch as the slider radius fixed to the base.Thus as the slider rotates, the band drive gear pushes theband at the same surface speed in the opposite direction sothe growing tube does not spin.

Fig. 4. Top view of the slider with gearing.

In addition, the band containment gear (Figure 3 whichis the outer planetary gear shown in Figure 4 rotates thebase housing the excess band material. This ensures that theexcess band does not tighten and bind or jumble and snagwhile extending or collapsing (resp.).

B. Adding Rotational Degrees of Freedom

By mounting the Spiral Zipper onto a gimbal (Figure 1)or universal joint, we form an RRP robot arm configuration.

In this configuration the position of the end point can becontrolled in a spherical coordinate form, as actuation andmeasurement are conveniently formed as pitch, roll, andradius. A full six DOF robot arm could be created by addinga wrist with 3 rotational DOF at the distal end.

The resulting system has a workspace that can be delimitedas a portion of a sphere. The outer radius of the sphere isthe fully extended length of the Spiral Zipper Ro ≈ H .There is an inner unreachable volume delimited by the fullycollapsed length of the Spiral Zipper Ri ≈ 2HB . The otherlimitations of the workspace are defined by the range ofmotion of the rotational joints. For example a universal jointwith both DOF having range of 180o would yield a hollowhemispherical workspace of radius Ro.

One nice characteristic of this workspace is that all pointsinterior to this workspace are singularity free which notablydoes not occur with most standard articulated robot arms.

As explained in Section IV, the band works best incompression and may be relatively weak under moments.Instead of actuating the pitch and roll DOF with motors atthe axis of the base, we can instead add tethers to the endeffector and control position by varying the length of thetethers with winches. This is illustrated in Figure 5.

Fig. 5. A rendering of a Spiral Zipper mounted on a universal joint basewith winch and cables.

If the rotational DOFs in the base are unactuated, (i.e. freeto move), that end of the tube cannot support any moments. Ifthere are no moments at the distal end and a force is appliedwhere the tethers intersect (e.g. a load is being carried atthe end effector), then there can be no moments in the tube.As a result there are, conveniently, only pure compressionforces in the tube. This condition is similar to ideal staticallydeterminate truss structures (e.g. truss bridges and cranes).

Fig. 6. A schematic view of an arm cantilevered holding a load at the endeffector.

Figure 6 illustrates the condition of forces along the tethersand the tube when the beam is cantilevered with a payloadunder gravity represented by a force F . There are several

trade offs visible from this figure. The tethers experiencepure tension FT (cables can only support tension), whilethe tube experiences pure compression FC as required bythe pinned end conditions of the beam assuming a masslessbeam.

This compressive force can become large as the anglebetween the tethers and the tube becomes small (e.g. thelength of the tube H becomes large compared to the baselinedistance to the winches r or α→ π/2 or β → π/2). In fact,the compressive force is exactly the ratio of the tube lengthto baseline distance in this particular case, FC = F H

rIt can be seen that a smaller baseline yields larger com-

pressive forces on the tube. The tradeoff here is that given alimited compressive stress (for example a critical bucklingload of a column), larger cantilever loads require largerbaseline profiles and this becomes severe as the column getslarge. In addition, as the joint angles at the base α and βin Figure 6, get smaller, the tether’s ability to apply torquesabout the base gimbal become reduced and the mechanicaladvantage becomes zero when α = 0 or β = 0.

III. IMPLEMENTATION

The implementations can be broken into three parts, theSpiral Zipper band, the slider and the tether gimbal system.

A. The Spiral Zipper

Many prototypes of the Spiral Zipper were constructedhowever, two versions will be discussed, a 57mm diameterversion (large) and a 17mm diameter version (small). Fig-ure 7 shows the large column extending to full height. Thefull extent of the column made out of a 15m long plasticband was over 2.2m tall.

Fig. 7. The 57mm diameter Spiral Zipper starting to extend (left) and fullyextended (right)

Most of the components in the Spiral Zipper other thanthe band and fasteners are 3D printed including the slider,gears, top tube cap and winding base.

The large version was created first to explore the feasibilityof the design. The small one explored the scalability. Theoverall architecture of the design is basically the same asthe larger one with a few changes. The height of the bandwas reduced from 50.8mm to 25.4mm tall. At a 17mmdiameter, it has a more aggressive incline angle than the57mm diameter one - 11.26 and 7.11 degrees respectively.This is advantageous because it extends or collapses fastergiven a fixed rotation speed. Because of the smaller diameter,the drive gear is more easily located on the outside of thetube where as the large diameters drive gear is on the inside.

Another design decision is where the drive gear shouldmesh with the band. It was observed that the band sometimeswould bind in one direction of motion depending on whetherthe band was driven on the first or second wrap layer. Adual ganged drive gear (see Figure 8) was designed so thatthe gear meshed in the bottom two layers simultaneouslyalleviating this issue.

Fig. 8. Photo of the slider and drive mechanism for the 17mm diameterprototype

The prototypes bands have been made out of Acetal andABS plastic. The bands were cut out of 4 inch and 2inch wide, 0.02-0.035 inch thick strips. The plastics werechosen for flexibility, durability, low friction, and ability tobe cleanly laser cut.

Manufacturing the long strips of the plastic band econom-ically is important to the goal of making the system low cost.The two-inch wide plastic band is readily available in variouslengths on a roll, but the challenge is precisely patterning theteeth on the edges. An ideal process would be a roll-to-rollband cutting operation. Custom machinery may be requiredfor large quantities, but for prototypes, a laser cutter wasused to cut successive portions of the band roll. For that thechallenge is to hold the band flat and align the band in thelaser cutter while maintaining registration of the band oversuccessive refixturing and cutting operations. In our case, acustom jig solved both issues.

The shape of the meshing teeth has two functions. One isto ensure sliding circumferentially along the edges does not

occur (thus changing radius and making the teeth disengage)while still being able to assemble at the slider. The other is tomaintain engagement. Different shaped teeth can also effectthe allowable tolerance that will result in slop in the tube(either in radius or in tension). Note that in the proposedsystem where the tube can maintain a compressive force thetolerance for motions in tension does not become a concern.Figure 9 shows some of the different teeth profiles and sizesthat were tested.

Fig. 9. Three of the dozen prototyped bands.

The designs that engaged well in the slider during assem-bly were ones with the most gap between the teeth in thedirection parallel to the joint that didn’t jam in the slider.For example, the top band in Figure 9 didn’t work becauseit had large dovetail teeth whose corners would remain planarwhile the band curved and thus would catch on the slider.

Critical to the bands performance is keeping the teeth fromdisengaging (moving locally out of plane). A backing ribbonalong one edge can help to ensure engagement. The tradeoffsin implementation include the ribbon thickness. It needs tobe thin enough to ease entry in the slider and minimizedifferential stresses from the varied thickness of the band.It can’t be too thin to maintain structural integrity.

For the large diameter prototype the backing consistedof 0.08mm thick Mylar adhered with double sided tape.Proper assembly of these adhesive bands is also critical.Adhering the ribbon in a flat state would cause wrinklingand delamination as the band was wrapped, so adhesion ona curve was required (ideally with a curvature similar to thatin the wrapped state). The double-sided tape was appliedbefore laser cutting the teeth profile such that the adhesivewas removed from between the teeth. The Mylar was thenadhered to this tape.

The gear slots labeled in Figure 9 are used to drive theband in the slider. They are angled at the helix angle of thewrap such that they remain aligned with the axis of the tubewhen assembled (Figure 8). While a gear tooth is engagedin the slot, it actually will translate within this slot becauseof the feeding angle of the band. So the slots are taller thanthe thickness of the gear.

B. Tether-gimbal system

Figure 1 shows the smaller diameter Spiral Zippermounted on a gimbal base supported by two tethers. Forthe three DOF system the three actuators (two winches and

the Spiral Zipper actuator) are sufficient under gravity. Inthe configuration shown, gravity ensures that the tethers arealways in tension. For convenience in this implementation,the tether routing is coplanar with the gimbal axis supportedby an 80-20 frame. Dynamixel EX106 servos capable of10Nm torque (not shown) were mounted with 50mm diam-eter spools to drive the winches and are capable of applying420N of force along the tether with no load speeds of238mm/s.

One set of applications for this arm is mounting it ontounmanned vehicles exploiting the low mass of the system.The US Army is interested in mounting the arm on anoctorotor being developed at the Army Research Lab. In thiscase the frame of the vehicle would be used instead of the80-20 frame. A mass analysis for this application using theprototype elements shown in Figure 1 is shown in Table I. Athird winch is added similar to Figure 5 making the systemover-constrained, but ensuring stability without considerationof gravity.

TABLE IMASS BUDGET

Item mass [g]Band [1m reach] 107Slider structure 64Spiral Zipper servo 154Gimbal 421EX106 servo, (x3) [420N @ stall] 470Gripper budget (e.g. simple hook) 200Structure budget 200Total 1717

IV. STRENGTH ANALYSIS

One of the main advantages of this system is the strengthto weight ratio. That is the payload carrying capabilities ofthe arm versus its own weight. As shown in Section II-B,the payload depends highly on the compressive strength ofthe tube. In fact, since the servos and tethers can apply over400N of pulling force, the main limitation will often be thecompressive strength of the tube.

A. Compression Testing

The compressive load capacity of a slender beam istypically due to the critical buckling load PCR = 4πEI

L2

where E is the Young’s modulus and I is the area momentof inertia for the tube. A hollow cylinder is optimal in termsof maximizing I for a given amount of material and uniformminimum buckling axis. However, since the Spiral Zipper ismade up of a band with teeth (not a smooth cylindrical shell)local plate buckling may occur which is much more difficultto analyze. A more empirical analysis is warranted.

Compressions tests were conducted on an materials testingmachine, MTS Criterion Model 43. This band was tested atdifferent extension lengths up to the maximum testing lengththe MTS machine could accommodate - 850mm.

The small band with curved T-shaped teeth had an interest-ing failure mode. Compression forces bent the teeth outwardcausing all the teeth to interlace under pressure as shown in

Figure 10 (right). As a result the integrity of the tube wouldstill be intact, just with a somewhat smaller length.

Fig. 10. Compressive failure modes. The 57mm diameter band locallybuckled (left). The 17mm band teeth interlaced (right).

A 1.5m long, large diameter band was also tested. It wastoo large to fixture in the MTS machine so, it was testedusing free weights. The column supported a maximum of530N of weight before failure occurred. In the large band,the point of failure was delamination of the Mylar backing,which allowed one the teeth to slip inside the radius asshown in Figure 10. Once delaminated, the column wouldstill function if the band was retracted and extended againpast the failure point, but at a reduced capacity. The zippermast (and in fact clothing zippers) recover similarly afterfailure [10]. After the backing was compromised, the columnonly supported approximately 350N before failure occurredin the already compromised locations.

The results of these tests are summarized in Table II. Inthis table, we include a theoretical buckling load for an idealEuler buckling analysis case. It shows that up to an orderof magnitude greater load may be a rough upper bound forthe theoretical maximum compressive load of the geometrytested.

TABLE IISPIRAL ZIPPER COMPRESSION TESTS

Diameter [mm] 114 33 33 33Material Acetyl Acetyl Acetyl ABSBand thickness [m] 8.9E-04 5.1E-04 5.1E-04 1.0E-03Length [m] 1.50 0.85 0.64 0.24Moment of inertia, I 5.1E-07 7.1E-09 7.1E-09 1.4E-08Modulus, E [GPa] 2.90 2.90 2.90 3.00Radius of gyration 0.04 0.01 0.01 0.01Slenderness ratio (l/r) 37 73 55 21PCR @ full ext [N] 25914 1119 2007 28137Actual failure [N] 530 200 260 1345

V. DISCUSSION

A. Advantages

The mechanism offers many key advantages. The first isthe low profile exhibited by the large extension ratio. Wewere able to achieve a 14:1 extension ratio on the largeband. The novel feature of this mechanism is no additionaloverhead needed to increase this ratio. One only has to addmore band. Ratios of 100:1 should be feasible as long as thestiffness of the resulting tube is not compromised.

Another advantage is the strength to weight ratio. Itis difficult to find devices designed exactly for the samepurpose, however, we can compare a range of arms that weredesigned to be lightweight.

TABLE IIICOMPARISON OF STRENGTH TO WEIGHT RATIO

Robot arm Reach Fmax Marm Fmax/FM Ref.Keemink et al. 0.09m 5 N 190g 2.68 N/N [2]LWR 4* 0.79m 69N 8.1kg 0.86 N/N [11]WAM 4DOF 1.0m 40N 25kg 0.16 N/N [12]Spiral Zipper Sys. 0.85m 100N 705g 14.4 N/N

*estimated for first 3DOF

Table III lists several such arms. As a relative measure,we use the maximum force the arm can apply when fullyextended Fmax divided by the weight FM =Marm×g. Formost robot arms this number will be less than 1.0 and isusually less than 0.1 [13].

The Keemink arm is a small arm designed to be mountedon a small UAV. It is a parallel mechanism made with carbonfiber rods, small range of motion, small payload but alsovery light. The Kuka Lightweight Arm 4 (LWR4) is a highperformance arm specifically designed to have a near 1:1ratio between payload and weight[13]. The LWR4 has 7 DOFand the other systems only have 3 DOF, so to make thecomparison more fair, we consider only the first 3 links of theLWR4. The Barrett WAM arm has been mounted on mobilerobots and is also cable driven and high performance. Likethe Spiral Zipper arm, the majority of the mass is located atthe base. For the Spiral Zipper arm, the mass budget fromTable I is used. While this does not include the mass of aframe (as the other systems do), for the mobile manipulationapplication, it is likely that mounting to the mobile base canreplace a frame. The winch motors can each pull 420N so thepayload limitation comes from the structural limitations. It isbased on 1/2 the compressive strength of the tube, assuminga baseline r = 0.42m and the 33mm diameter tube shown incolumn 2 of Table II. This is conservative as later versions ofthe band made of ABS do not exhibit the same delaminationissues resulting in higher compressive loads.

Also to be fair, speed is often a concern and the LWR4 andWAM systems are estimated to have 5 and 10 times (resp.)faster than the proposed system, for tip speed at no load,based on published data sheets. The Keemink arm’s no loadspeed is approximately the same as the spiral zipper basedon the Maxon RE10 motor they use.

B. Concerns

There are a few disadvantages that are inherent to thedesign that need to be considered. Keeping long bands (e.g.15m) contained and orderly is a challenge. If the columnbreaks or the idle spool comes out, the band tends toget tangled or becomes very hard to work. Reassemblingthe column was made easier by the addition of an accessdoor on the small diameter prototype. Maintaining the baserotation speed with the containment gearing was surprisinglychallenging.

The 15m long band had major stability issues at thefully extended length. Each tooth has negligible play in theperpendicular direction to the joint in any one wrap; however,over many wraps, the play is accumulated. While still stableunder compression, the column would lean from side to side

causing large moments at the base. If the play is too large,the column teeth could also disengage at full extension.

The loading condition for a generic manipulator may notresult in a pure force at the intersection of the mountedtethers. There may either be an offset force, or the loadmay present some moments at the end effector. This maythen induce moments that must be supported by the SpiralZipper tube. Tighter toleranced teeth than those presented inthis paper will directly result in better performance undertension and moments. Later versions with tolerances of0.02mm on the teeth significantly reduced this play, but theaccumulation will still become an issue at long extensionsand more structured testing under moment conditions needsto be developed.

VI. CONCLUSIONS

We were able to demonstrate promising prototypes for thisSpiral Zipper based spherical robot and look to improve onthe current performance of the column extension mechanism.As a robot arm mounted on a mobile base, having extensionratios of greater than 10 eases maneuvering and manipulationin cluttered environments.

For high strength to weight ratio actuation over longdistances, the proposed system is typically an order ofmagnitude higher performing than other robot arm systemsdesigned to be lightweight. Further improvements such asoptimizing motors and servo’s for specific torque as wellas improving the backing tape of the bands for highercompressive strength is likely to greatly improve this ratioas well.

We are currently working on a possible applicationsmounting the arm on mobile bases including indoor aids forthe elderly, and mounting to an octorotor for applicationssuch as opening doors and manipulation of small objects.Combining the large reach with strong winch motors mayalso enable the manipulation of large objects such as furni-ture or spring loaded doors with indoor mobile robots.

ACKNOWLEDGEMENT

The core technology presented in this work has been patented [5] and isbeing licensed by Prendo Systems Inc. The authors would like to thank RuiZhang for his help in implementing the tether prototypes. This work wasfunded in part by NSF award IIP-1430216 and Army RCTA program.

REFERENCES

[1] S. S. Srinivasa, D. Ferguson, C. J. Helfrich, D. Berenson, A. Collet,R. Diankov, G. Gallagher, G. Hollinger, J. Kuffner, and M. V. Weghe,“Herb: a home exploring robotic butler,” Autonomous Robots, vol. 28,no. 1, pp. 5–20, 2010.

[2] A. Keemink, M. Fumagalli, S. Stramigioli, and R. Carloni, “Mechani-cal design of a manipulation system for unmanned aerial vehicles,” inRobotics and Automation (ICRA), 2012 IEEE International Conferenceon. IEEE, 2012, pp. 3147–3152.

[3] G. S. Hegde, A Texbook of Industrial Robotics. Laxmi Publications,2006.

[4] D. Reznik and V. Lumelsky, “Sensor-based motion planning in threedimensions for a highly redundant snake robot,” Advanced robotics,vol. 9, no. 3, pp. 255–280, 1994.

[5] M. Yim, “Linear motion device with extending tube for positioning,”January 2014, patent App. PCT/US2015/012609.

[6] J. Albus, R. Bostelman, and N. Dagalakis, “The NIST robocrane,”Journal of Robotic Systems, vol. 10, no. 5, pp. 709–724, 1993.

[7] E. Grassmann, “Lifting jack,” Sept. 19 1939, uS Patent 2,173,685.[8] I. Geo Systems, “Zm-20 zippermast,” http://zippermast.com/sam-15/,

accessed: 2015-09-01.[9] Paco Group, “Spiralift, how it works,”

http://www.pacospiralift.com/howitworks.htm, 2011, accessed:2015-09-01.

[10] G. Woodruff, P. Muench, and G. Witus, “Zipper mast for enhancedcommunications and surveillance,” in SPIE Defense, Security, andSensing. International Society for Optics and Photonics, 2011, pp.804 512–804 512.

[11] C. Gaz, F. Flacco, and A. De Luca, “Identifying the dynamic modelused by the KUKA LWR: A reverse engineering approach,” inRobotics and Automation (ICRA), 2014 IEEE International Conferenceon. IEEE, 2014, pp. 1386–1392.

[12] Barrett Technology Inc., “WAM Arm data sheet,”http://www.barrett.com/DS WAM.pdf, 2011, accessed: 2015-09-01.

[13] R. Bischoff, J. Kurth, G. Schreiber, R. Koeppe, A. Albu-Schaffer,A. Beyer, O. Eiberger, S. Haddadin, A. Stemmer, G. Grunwald, et al.,“The kuka-dlr lightweight robot arm-a new reference platform forrobotics research and manufacturing,” in Robotics (ISR), 2010 41stinternational symposium on and 2010 6th German conference onrobotics (ROBOTIK). VDE, 2010, pp. 1–8.